This article presents an intervention proposal that seeks to develop computational thinking in Early Childhood Education, using an educational ground robot with programmed directionality controls. Within the observational methodology, it has been possible to characterize computational thinking in children of Early Childhood Education -- 5 years old, during the resolution of each of the problems of increasing difficulty that constitute the intervention proposal. The data was recorded and coded using the LINCE software. The reliability of the data, calculated in the form of inter-observer agreement using Cohen's Kappa coefficient, has been guaranteed. The analysis of generalizability carried out allows to guarantee the homogeneity in the behavior of the participants. Two techniques of diachronic analysis of observational data have been applied in a complementary way: the lag sequential analysis, with the free software GSEQ, and the analysis of polar coordinates, with the free software HOISAN. Skills associated with computational thinking such as the logical organization of the steps to follow, the physical-concrete abstraction and generalization of the trajectories, and the self-evaluation of the proposals, have been conditioned by the development of the capacity for spatial organization of the 5 years old child and by the intervention of the teacher.